Search Abstracts | Symposia | Slide Sessions | Poster Sessions
From Pitch to Meaning: sEEG Evidence from Novel Lexical Tone Learning
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Yixiang Liu1, Hao Zhu1, Chen Yao2, Junxi Chen3, Xing Tian4, Xiangbin Teng1, Patrick C.M. Wong1; 1The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China, 2Department of Neurosurgery, Shenzhen Second People’s Hospital, The First Affiliated Hospital of Shenzhen University, Shenzhen, Guangdong, China, 3Epilepsy Center, Guangdong Sanjiu Brain Hospital, Guangzhou, Guangdong, China, 4NYU-ECNU Institute of Brain and Cognitive Science at NYU Shanghai, Shanghai, China
Prior functional magnetic resonance imaging (fMRI) research shows that tone-language experience facilitates novel lexical tone learning, engaging the superior temporal gyrus (STG), inferior frontal gyrus (IFG), and corticostriatal circuits. However, the blood-oxygen-level-dependent (BOLD) signal lacks the fine temporal resolution to dissociate acoustic from semantic encoding and provides only indirect neurophysiological evidence. Electrocorticography (ECoG) provides direct cortical electrophysiological recordings, but it has primarily been used to investigate the neural processes underlying established tone perception or acquisition of pitch categories, rather than learning of novel tone-meaning associations. Meanwhile, ECoG cannot access the hippocampus, which is essential for forming new associative memories. Therefore, we designed a 3-day learning paradigm in which 16 Mandarin-speaking patients with drug-resistant epilepsy learned to map novel tonal patterns (found in Thai) onto pseudowords for lexical meaning during stereoelectroencephalography (sEEG) recordings. Each day, participants completed two sessions of a three-stage task: exposure (encoding of sound-meaning associations), retrieval practice with feedback, and testing without feedback. Participants' performance improved significantly across the three days. 221 auditory-responsive contacts showing significant post-onset high-gamma activity (HGA; 70-150 Hz) were identified. For each responsive contact, we computed stimulus-wise HGA patterns and constructed representational dissimilarity matrices (RDMs) across the 15 items (5 tones × 3 syllables) per session. Despite prior fMRI evidence for contrast activation in the left STG, sEEG showed increased representational separability after training, with larger RDM changes farther from the right mid-STG and no comparable gradient in the left STG. Unique variance partitioning of session-wise RDMs identified dissociable contact groups whose representational separability tracked exposure, accuracy, or shared learning-related variance. In learning-effect contacts, session-wise RDMs correlated positively with performance during the exam phase. Further, neural RDMs were compared with feature-wise pitch-contour and semantic-embedding distance matrices. Pitch-tuning contacts were restricted to the right hemisphere, spanning STG, Heschl’s gyrus, inferior frontal gyrus, putamen, insula, precentral gyrus, and postcentral gyrus, which showed increased selectivity for pitch differences over learning, while semantic-tuning and stable semantic-encoding contacts were distributed bilaterally. To investigate potential subcortical contributions to the learning process, we next examined hippocampal ripples, high-frequency events associated with the formation and retrieval of newly acquired associations. Ripple rates showed right-lateralized anterior-posterior differences during auditory and image presentation and were higher for incorrect than correct retrieval trials, consistent with greater retrieval demand when tone-meaning retrieval failed. Together, sEEG addressed the limitations of previous studies by combining millisecond-level direct neurophysiological signals with simultaneous cortical-subcortical recordings. These results revealed a fronto-temporal network for phonology-semantics encoding during novel lexical tone learning, which is supported by hippocampal-mediated retrieval.
Topic Areas: Language Development/Acquisition,